一种新的区间模糊数时间序列预测模型

The BP neural network based on GA optimization in interval fuzzy number time series prediction

通过对区间模糊数进行研究分析,发现已有处理方法是直接对两个界点建模和预测,这样做往往会导致不能很好的描述序列整体性的发展趋势以及模型所预测的结果容易发生错乱等,从而预测失效.首先基于等价和整体性考虑提出了模糊序列的面积序列和重心序列概念.然后对面积序列和重心序列分别建立了遗传优化BP神经网络模型进行回归和预测,并通过还原公式得到原区间模糊数序列的拟合值和预测值.最后通过实例验证了该方法有效可行,对比传统的BP神经网络模型和ARIMA模型,显著提高了预测精度.

Through the research and analysis of interval fuzzy number,it is found that the existing processing method is to directly model and predict the two boundary points,which often leads to the defects of not well describing the development trend of integrity and the results predicted by the model are prone to be disordered,so as to predict the failure.Firstly,the concepts of area sequence and center of gravity sequence of fuzzy sequences are proposed based on equivalence and integrity.Then,the BP neural network model of genetic optimization is established for regression and prediction of area sequence and center of gravity sequence respectively,and the fitting value and prediction value of original interval fuzzy number sequence are obtained by reduction formula.Finally,example is given to verify the effectiveness and feasibility of this method.Compared with the traditional BP neural network model and ARIMA model,the prediction accuracy is significantly improved.